OPEN SOURCE

FIBEMATE v3.3 — 开源发布公告

2026-08-31 · GPL-3.0 · GitHub

FIBEMATE is now open source. 🎉
What does it take to ship post-quantum cryptography end-to-end — from browser to FPGA, from NIST standards to Chinese national algorithms, from KAT vectors to TLA+ formal proofs?

Two years. One person. 100 timestamped evidence records. Here's what we found.

What FIBEMATE Is

A post-quantum cryptography executable textbook. Not a security product — an engineering proof that the full chain works.

Track	Coverage	Status
Standard PQC	ML-KEM-768 (FIPS 203), NTT domain, C Native addon (15-31× speedup), WASM	✅ Verified
National Crypto Hybrid	SM2/SM3/SM4-GCM + ML-KEM-768, IANA #4590, double-ratchet PQ mix-net	✅ Verified
Hardware (FPGA)	NTT pipeline on Artix-7 35T, UART physical debug passed, 256×256 roundtrip	✅ Live
Protocol Verification	TLA+ Path C-2 model, 7 invariants, 101,467 states, K3 strong key independence	✅ Formal
Research (default-off)	VWZ tensor signatures, LookingGlass v2 — isolated branch, zero security claims	⚠️ Lab

Why This Exists

The post-quantum migration is an engineering problem, not a paper problem.

FIPS 203 exists. liboqs exists. But stitching them together — WebCrypto ↔ FPGA BRAM, SM2 legacy ↔ ML-KEM hybrids, unit tests ↔ TLA+ models — that's where the real work lives.

FIBEMATE answers: Can one person build, verify, and timestamp-evidence the entire stack? Yes. With honest caveats.

What Works (verified with evidence)

Core Cryptography

• ML-KEM-768: NTT domain (FIPS 203 aligned) · noble 10K/10K ✅ · liboqs 10K/10K ✅
• C Native Addon: keygen 103µs (15.8× JS), encaps 92µs (31.5×), decaps 184µs (14.1×)
• SM2: GB/T 32918 · constant-time Barrett reduction (14×) · scalar masking · software TVLA N=10K PASS
• SM3/SM4-GCM: KAT vectors · pure JS benchmarks (SM3 21K ops/s, SM4-GCM ~230KB/s)
• Hybrid KEX: SM2 + ML-KEM-768 · IANA #4590 registration
• Double Ratchet PQ: ML-KEM-768 + P-256 hybrid handshake → multi-round message encryption ✅

Protocol Verification

• TLA+ Path C-2: 7 invariants all pass, TLC EXIT 0, 101,467 states
• K3 strong key independence: formally proven ∀ i≠j: key[i]≠key[j]

FPGA (Artix-7 35T)

• NTT 256×256 roundtrip verified on physical hardware
• UART debug channel operational (CH340 external adapter)
• Timing: WNS 9.869ns, WHS 0.131ns — plenty of slack

Reproducibility

• 100 RFC 3161 TSR records (lg-001~lg-100) — DigiCert + FreeTSA dual authority
• Every KAT, cross-validated, and release milestone timestamped
• TSR manifest: v4, 131 total entries, SHA256 inventory
• CI: Nightly ✅ | node-test ✅ | cross-lang ✅ | KAT smoke ✅

What's NOT Claimed (read this)

Claim	Reality
Security product	❌ Engineering demonstration platform
Third-party audit	❌ Self-tested, cross-validated, not externally reviewed
Novel cryptography	❌ VWZ/LG are default-off research, zero security guarantees
Hardware security	❌ No secure element, physical TVLA pending Q4 2026
ML-KEM formal proof	❌ TLA+ covers handshake protocol, not lattice math

Quick Start (30 seconds)

git clone https://github.com/Lennonhaha/fibemate.git && cd fibemate && npm ci
node -e "
  const { generateKeypair, encapsulate, decapsulate } = require('./packages/pqc-kem');
  const kp = generateKeypair();
  const { ciphertext, sharedSecret } = encapsulate(kp.publicKey);
  const ss = decapsulate(kp.secretKey, ciphertext);
  console.log(ss.equals(sharedSecret) ? 'ML-KEM-768: PASS' : 'FAIL');
"

Live demo: fibemate.net/demo

Who This Is For

• Crypto engineers → wire format, cross-validation, side-channel mitigations
• Security researchers → TLA+ model, Path C-2 hybrid design, failure modes
• FPGA developers → NTT pipeline, BRAM layout, behavioral model
• Students & learners → architecture docs cover every layer, browser to silicon

The Numbers

Metric
ML-KEM-768 Native keygen	103 µs
ML-KEM cross-validation	20,000/20,000 (noble + liboqs)
NTT Barrett reduction speedup	14×
SM2 software TVLA	36/36 PASS (N=10,000)
ML-KEM-768 TVLA	corrected (V8 JIT artifact, no exploitable leak)
ML-KEM-1024 TVLA	3/3 PASS (Noble, N=10,000)
TLA+ verified states	101,467
TSR evidence records	131 total (lg-001~lg-100)
CI badges	✅ all green
Project score	9.3/10

Deep-Dive Series (coming weekly after launch)

Week	Topic
1	Double Ratchet PQ — ML-KEM + P-256 hybrid handshake design
2	FPGA NTT Pipeline — from behavioral model to UART debug
3	SM2 Repair Log — _fastModP dead loop, wNAF timing leak, 0.2% edge case
4	TSR Evidence Chain — 100 RFC 3161 timestamps, why they matter
5	TLA+ Path C-2 — formal verification of hybrid key exchange

How to Contribute

Solo project opening up. Here's how:

• First issues → good-first-issues
• Cross-platform testing → Windows/macOS/Firefox/Safari matrix incomplete
• Security review → found a bug? SECURITY.md
• Hardware → got a ChipWhisperer? FPGA needs physical TVLA

"The post-quantum transition is not about swapping algorithms. It's about proving the chain holds."
— FIBEMATE, v3.3, 2026-08-31